Topology: Weak and Weak* Topologies, Study notes of Mathematics

The concepts of weak and weak* topologies in the context of real vector spaces and their duals. It covers the definitions, properties, and relationships between these topologies, including tychonoff's product topology, the topology of pointwise convergence, and the banach-alaoglu theorem. The document also includes proofs for various propositions and lemmas.

Typology: Study notes

2010/2011

Uploaded on 09/09/2011

andreasphd
andreasphd 🇬🇧

4.7

(28)

287 documents

1 / 3

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
5. Weak and weak* topologies
Definition 5.1. Let Zbe a set and let (Zi,Ti)iIbe a family of topological spaces. For each ilet pi:ZZibe a mapping. We
define the initial topology on Z(associated with the mappings pi) to be the coarsest topology on Zfor which all the mappings pi
are continuous. Its basic open sets are of the form
\
iF
p1
i[Ui],
with Fa finite subset of Iand UiTifor each i.
There are many important examples. For instance, given the family (Zi)iIas above, we may consider the product space
Z=Y
iI
Zi,
whose elements are all functions fthat assign to each iIan element f(i) of Zi. (We sometimes prefer to use “family” notation
(zi)iIfor such things, in order to emphasize the similarity with sequences and the idea that f(i) is the ith co-ordinate” of f.)
If all the spaces Ziare copies of a single space Z, we write ZIfor the product space. On the product space Zwe introduce the
co-ordinate mappings
πi:ZZi;f7→ f(i),
and define the (Tychonoff) product topology to be the initial topology with respect to the mappings πi. It is an important theorem
(Tychonoff’s Theorem) that if all the spaces Ziare compact and Hausdorff, then so is Z. This theorem is presented in the C1
Analytic Topology course.
If Lis a topological space then the space of all continuous real-valued functions C(L) is a subspace of RL. The topology induced
by the product topology is the topology of pointwise convergence, basis open neighbourhoods of a function fbeing of the form
{yC(L) : |y(ti)x(ti)|< for i= 1,2, . . . , n},
where nis a natural number, > 0 and t1, . . . , tnL.
Let Xbe a real vector space and let Ybe a subspace of the algebraic dual X#. We define the topology σ(X, Y ) to be the
coarsest topology on Xfor which all the maps g:XR(gY) are continuous. Basis open sets have the form
{x0X:|gi(x0)gi(x)|< for i= 1,2, . . . , n}.
In the important case where Xis a normed space and Y=Xwe call this topology the weak topology on X.
When Xis a normed space, we can also introduce a topology on X, defined to be the coarsest topology for which all the
mappings f7→ f(x) (xX) are continuous. This is called the weak* topology on X. In the σ-notation, this topology is just
σ(X, JX[X]) (or, less pedantically, σ(X, X). Notice that the weak topology on Xis coarser than the norm topology (because
all fXare norm-continuous). In Proposition 5.4 we show that the weak topology is strictly coarser than the norm topology
whenever Xis infinite-dimensional. Note also that the weak* topology on Xis coarser (and in the non-reflexive case strictly
coarser) than the weak topology σ(X, X∗∗ ).
Lemma 5.2. If Xis a real vector space and Yis a subspace of X#then every σ(X,Y )-open neighbourhood of 0contains a
finite-codimensional subspace X1of X.
Proof. A basic σ(X, Y )-open neighbourhood of 0 is of the form
{xX:|gj(x)|< ifor i= 1,2, . . . , n}
where giYand i>0. This set contains the finite-codimensional subspace
X1=\
in
ker gi.
Proposition 5.3. If Xis a real vector space and Yis a subspace of X#then a linear mapping g:XRis σ(X, Y )-continuous
if and only if gY.
Proof. If gYthen gis σ(X, Y )-continuous by definition of the topology. On the other hand, if gis linear and σ(X , Y )-continuous
then there exists a σ(X, Y )-open neighbourhood Uof 0 such that |g(x)| 1 for all xU. It follows that gis zero on the subspace
X1=Tinker giof Lemma 5.2. It follows “by algebra” that gis a linear combination Pinλigiand hence in the subspace Y.
Proposition 5.4. If Xis an infinite-dimensional normed space then the weak topology on Xis strictly coarser than the norm
topology.
Proof. Each fXis continuous for the norm topology, and so this topology if finer than the weak topology (by the definition of
σ(X, X ) as the coarsest topology making all fXcontinuous.
The unit sphere SX={xX:kxk= 1}is closed in the norm topology. I shall show that SXis not closed in the weak topology
whenever Xis infinite-dimensional. By Lemma 5.2, every weak neighbourhood Uof 0 contains a subspace X1of Xwhich is of
27
pf3

Partial preview of the text

Download Topology: Weak and Weak* Topologies and more Study notes Mathematics in PDF only on Docsity!

  1. Weak and weak* topologies

Definition 5.1. Let Z be a set and let (Zi, Ti)i∈I be a family of topological spaces. For each i let pi : Z → Zi be a mapping. We define the initial topology on Z (associated with the mappings pi) to be the coarsest topology on Z for which all the mappings pi are continuous. Its basic open sets are of the form (^) ⋂

i∈F

p− i 1 [Ui],

with F a finite subset of I and Ui ∈ Ti for each i. There are many important examples. For instance, given the family (Zi)i∈I as above, we may consider the product space

Z =

i∈I

Zi,

whose elements are all functions f that assign to each i ∈ I an element f (i) of Zi. (We sometimes prefer to use “family” notation (zi)i∈I for such things, in order to emphasize the similarity with sequences and the idea that f (i) is the “ith^ co-ordinate” of f .) If all the spaces Zi are copies of a single space Z, we write ZI^ for the product space. On the product space Z we introduce the co-ordinate mappings πi : Z → Zi; f 7 → f (i),

and define the (Tychonoff) product topology to be the initial topology with respect to the mappings πi. It is an important theorem (Tychonoff’s Theorem) that if all the spaces Zi are compact and Hausdorff, then so is Z. This theorem is presented in the C Analytic Topology course. If L is a topological space then the space of all continuous real-valued functions C (L) is a subspace of RL. The topology induced by the product topology is the topology of pointwise convergence, basis open neighbourhoods of a function f being of the form

{y ∈ C (L) : |y(ti) − x(ti)| <  for i = 1, 2 ,... , n},

where n is a natural number,  > 0 and t 1 ,... , tn ∈ L. Let X be a real vector space and let Y be a subspace of the algebraic dual X#. We define the topology σ(X, Y ) to be the coarsest topology on X for which all the maps g : X → R (g ∈ Y ) are continuous. Basis open sets have the form

{x′^ ∈ X : |gi(x′) − gi(x)| <  for i = 1, 2 ,... , n}. In the important case where X is a normed space and Y = X∗^ we call this topology the weak topology on X. When X is a normed space, we can also introduce a topology on X∗, defined to be the coarsest topology for which all the mappings f 7 → f (x) (x ∈ X) are continuous. This is called the weak* topology on X∗. In the σ-notation, this topology is just σ(X∗, JX [X]) (or, less pedantically, σ(X∗, X). Notice that the weak topology on X is coarser than the norm topology (because all f ∈ X∗^ are norm-continuous). In Proposition 5.4 we show that the weak topology is strictly coarser than the norm topology whenever X is infinite-dimensional. Note also that the weak* topology on X∗^ is coarser (and in the non-reflexive case strictly coarser) than the weak topology σ(X∗, X∗∗).

Lemma 5.2. If X is a real vector space and Y is a subspace of X#^ then every σ(X, Y )-open neighbourhood of 0 contains a finite-codimensional subspace X 1 of X.

Proof. A basic σ(X, Y )-open neighbourhood of 0 is of the form

{x ∈ X : |gj (x)| < i for i = 1, 2 ,... , n}

where gi ∈ Y and i > 0. This set contains the finite-codimensional subspace

X 1 =

i≤n

ker gi.

Proposition 5.3. If X is a real vector space and Y is a subspace of X#^ then a linear mapping g : X → R is σ(X, Y )-continuous if and only if g ∈ Y.

Proof. If g ∈ Y then g is σ(X, Y )-continuous by definition of the topology. On the other hand, if g is linear and σ(X, Y )-continuous then there exists a σ(X, Y )-open neighbourhood U of 0 such that |g(x)| ≤ 1 for all x ∈ U. It follows that g is zero on the subspace X 1 =

i≤n ker^ gi^ of Lemma 5.2. It follows “by algebra” that^ g^ is a linear combination^

i≤n λigi^ and hence in the subspace^ Y^.^ 

Proposition 5.4. If X is an infinite-dimensional normed space then the weak topology on X is strictly coarser than the norm topology.

Proof. Each f ∈ X∗^ is continuous for the norm topology, and so this topology if finer than the weak topology (by the definition of σ(X, X∗) as the coarsest topology making all f ∈ X∗^ continuous. The unit sphere SX = {x ∈ X : ‖x‖ = 1} is closed in the norm topology. I shall show that SX is not closed in the weak topology whenever X is infinite-dimensional. By Lemma 5.2, every weak neighbourhood U of 0 contains a subspace X 1 of X which is of

27

28

finite codimension, and hence non-trivial. Thus U ∩ SX ⊇ X 1 ∩ X = SX 1 6 = ∅. This shows that 0 ∈ SX weak , so that SX is not weakly closed. 

Theorem 5.5 (Banach–Alaoglu). If X is a normed space, then the dual unit ball BX∗ is weak*-comnpact.

Proof. A functional f ∈ BX∗^ is a (linear) mapping with the property that f (x) ∈ [−‖x‖, ‖x‖] for each x ∈ X. So we can regard BX∗^ as being a subset of the product space K =

x∈X [−‖x‖,^ ‖x‖]. [Recall that the elements of^ K^ are^ all^ functions^ f^ on^ X^ such that |f (x)| ≤ ‖x‖ for all x ∈ X.] Moreover, the topology induced on BX∗^ by the product topology on K is exactly the weak* topology on BX∗^ [in each case, we are looking at the coarsest topology that makes all the maps f 7 → f (x) continuous]. Since each [−‖x‖, ‖x‖] is a compact Hausdorff space, so too is K, by Tychonoff’s theorem. To prove our theorem it is enough, therefore, to show that BX∗^ is a closed subset of K. Now, for every x ∈ X the map f 7 → f (x) is continuous on K; so if x 1 , x 2 ∈ X and λ ∈ R the mapping qx 1 ,x 2 ,λ : K → R : f 7 → f (x 1 + λx 2 ) − f (x 1 ) − λf (x 2 )

is continuous on K. Hence the intersection (^) ⋂

x 1 ,x 2 ∈X; λ∈R

q− x 11 ,x 2 ,λ(0),

which is exactly BX∗^ , is closed in K. 

Corollary 5.6. For every normed space X there exists a compact Hausdorff space K and a linear isometric embedding X → C (K).

Proof. Take K = BX∗ in the weak* topology. For each x ∈ X the function Jx : K → R; f 7 → f (x) is continuous on K (by definition of the weak* topology), and ‖Jx‖∞ = sup f ∈BX∗

|f (x)| = ‖x‖,

by a standard corollary to the Hahn–Banach theorem. 

Theorem 5.7 (Non-examinable). For a normed space X, the dual ball BX∗ is weak* metrizable if and only if X is separable.

Proof. One implication (X separable =⇒ BX∗^ metrizable) has been set as a problem. The converse can be deduced from the fact that C (K) is a separable Banach space whenever K is compact and metrizable. This in turn follows from the Weierstrass–Stone Theorem (not on this course), or from a direct argument involving partitions of unity which we shall not give here. 

It follows that every separable normed space embeds into C (K) for some compact metric space K. One can do even better: every separable normed space embeds isometrically into C [0, 1]. The proof of this refinement is related to the “Hahn–Mazurkiewicz” theorem of General Topology.

Proposition 5.8 (A variant of the H–B Separation Theorem). Let Y be a normed space, let C be a convex subset of the dual space Y ∗^ and let g ∈ Y ∗. If g is not in the weak*-closure of C then there exists y ∈ Y such that

g(y) > sup f ∈C

f (y).

Proof. As in other separation theorems, we may assume that 0 ∈ C. If g /∈ C w* then there exists a weak* open neighborhood U of g such that U ∩ C = ∅. Now U has the form

U = {f ∈ Y ∗^ : |f (yi) − g(yi)| < i for i = 1, 2 ,... , n} = g + V,

where V is the weak* open neighborhood of 0 given by

V = {h ∈ Y ∗^ : |h(yi)| < i for i = 1, 2 ,... , n}.

The set C˜ = C + V is open and g /∈ C˜, so (as in the HBST proved earlier) there exists η ∈ Y ∗∗^ such that

η(g) ≥ sup f ∈C+V

η(f ) > sup f ∈C

η(f ).

Since V ⊇

i≤n ker^ JY^ (yi) we have (“by algebra” as in Proposition 5.3)^ η^ ∈^ sp〈JY^ (yi) :^ i^ ≤^ n〉^ whence^ η^ =^ JY^ (y) for some y ∈ Y X. 

Theorem 5.9 (Goldstine). If X is a normed space then JX [BX ] is weak*-dense in BX∗∗^.

Proof. Let X be a normed space and let g ∈ X∗∗^ \ JX [BX ]

w* ; we shall show that ‖g‖ > 1. Taking Y = X∗^ and C = JX [BX ], we deduce from Proposition 5.8 that there exists f ∈ X∗^ such that

g(f ) > sup{(JX x)(f ) : x ∈ BX } = sup{f (x) : x ∈ BX } = ‖f ‖.

This implies that ‖g‖ > 1 

Theorem 5.10. A Banach space X is reflexive if and only if BX is compact in the weak topology.